# Is diffraction an increasing worry? Or, what's wrong with more megapixels?

Started 3 months ago | Question thread
 Like?
 Re: m4/3s becomes diffraction limited at ... In reply to Anders W, 3 months ago

Anders W wrote:

Detail Man wrote:

richarddd wrote:

Sanpaku wrote:

According to the Airy disk calculator at Cambridge in Colour, 4/3 and m4/3 sensors become diffraction limited at the following apertures:

You might want to read this post http://forums.dpreview.com/forums/post/50918646, including "Please would people stop linking to that flawed, misleading article. The 'diffraction limit' calculated by that calculator is something that does not exist. People get obsessed by trying to avoid something that doesn't exist. Diffraction exists, but a pixel-size based 'diffraction limit' is a nonsense."

Yes, an absolute "diffraction limit" in the sense of an abrupt "extinction" of resolution it is not. "Degradation" would be a better term. If it bothers you that your M43 camera system that resolves 16 Mpixels (Bayer-arrayed) at F=5.6 "degrades" to a 4 Mpixel camera system at F=11, and to a 1 Mpixel camera system at f=22, then you (might, perhaps) find that situation less than satisfying.

http://forums.dpreview.com/forums/post/50923531

He did, opining that it was "a very nice post". Point is, it's accurate to say that "degradation" exists.

While we agree, of course, that the impact of diffraction is gradual, the MFT illustration you offer, suggesting that image resolution effectively halves with every two-stop reduction of aperture from f/5.6 on, is unrealistically stark.

If the diameter of a circle of light equals the size of 2x2 photosites, we call it (Bayer-arrayed) 16 Mpixels. If the diameter of a circle of light equals the size of 4x4 photosites, I call it (Bayer-arrayed) 4 Mpixels. If the diameter of a circle of light equals the size of 16x16 photosites, I call it (Bayer-arrayed) 1 Mpixels. Is that "bone-headed" thinking ? What am I doing wrong there ?

Are you talking about light that exists outside of the diameter of such circles (yet inside the square-shaped areas of the NxN photo-site arrays that I am describing) ? Assuming an (unlikely) 100% Fill Factor, that represents (at most) 21.46% of the square area. Is that the "missing link" ?

Note that image resolution, i, relates to lens resolution, l, and sensor resolution, s, approximately as follows:

1/i = sqrt(1/l^2 + 1/s^2)

where the value of l of course depends on the lens as well as the aperture used and does so not only as a result of diffraction but also lens aberrations and the extent to which they are reduced by stopping down.

In practice, the reduction of image resolution between f/5.6 and f/11 is, on average, far less than 50 percent and it does not quite reach 50 percent between f/11 and f/22 either.

I have seen that identity mentioned in various writings. I have also seen it stated many times that such an identity represents only a rough and often not representative approximation of what occurs. Am curious as to the nature of the derivation of that identity. Do you have any related information or references that derive and show the nature and validity of that identity ? Am sincerely interested.

Edited 3 months ago by Detail Man
Complain
Keyboard shortcuts: